Investigations into deforestation, forest fragmentation and the distribution of threatened plant species in the South-east of Brazil

Abstract

Land use and cover change are resulting in changes across the globe, and especially in tropical ecosystems. Investigating factors that relate to these disturbances is essential if we are to mitigate the damage they are causing to biodiversity and ecosystem services. This is especially important in the case of megadiverse and threatened tropical biomes, such as the Brazilian Atlantic Forest, considered a biodiversity hotspot. The relationships between the greatest threats to the Atlantic Forest – deforestation and forest fragmentation – and external factors that might be related to them are not currently well understood, and comprehending them is vital for the conservation of the forest’s remaining area. Equally important is identification of priority areas that need protection. Understanding the potential distribution of threatened species is a key tool for identifying those areas. This thesis aims to investigate the relationships between deforestation and forest fragmentation, and socio-economic and bio-geophysical factors in the Atlantic Forest in the state of Minas Gerais, Brazil, and to model the potential distributions of red-listed plant species, which occur in this biome, and Cerrado and Caatinga of Minas Gerais. In Chapter 2 I used random forest analysis (RF) to identify relationships between socio-economic and bio-geophysical factors, and deforestation and forest fragmentation metrics in the Atlantic Forest. In chapter 3, I used RF to investigate how those relationships changed when they were analysed at different spatial scales. In Chapter 4, I used MaxEnt to investigate the potential distribution of redlisted plant species in Minas Gerais, in order to recommend priority areas for conservation based on the highest environmental suitability for the species modelled. Two broader-scale projects carried out in Minas Gerais – the ecologiceconomical zoning of Minas Gerais, and the vegetation monitoring system dataset – provided the factors, which number more than 300, deforestation and forest fragmentation metrics, and some of the occurrence points of threatened plant species. I compiled further occurrence points from the SpeciesLink Website, and NeoTropTree database. RF analysis proved to be an effective tool for elucidating the relationships between factors and metrics, both in comparison to the classical approach and when used at multiple scales. Overall, a set of factors from different categories was identified as having relatively strong relationships with patterns of deforestation and forest fragmentation. Road density was the most commonly selected factor in the RF models. I observed extensive variation in the RF models at different scales, with some metrics providing higher levels of explanation at finer scales, while others at larger scales. Additionally, I have shown that deforestation processes are still occurring in the Atlantic Forest of Minas Gerais, even with specific legislation that makes deforestation in this biome illegal. Finally, I identified significant gaps in the protection of threatened plant species, mainly in the Atlantic Forest. Together with areas identified in the Cerrado and Caatinga, these must be considered as priorities for conducting inventories and, after confirmation of existing species distributions, proposing new natural reserves.